Elon Musk frequently speaks up for China's AI, yet the truth has been overlooked.

While everyone was discussing the sudden departure of Lin Junyang, the person in charge of Alibaba's Qianwen large model, another event was overlooked.

Just two days ago, Lin Junyang had an interaction with Elon Musk. This interaction stemmed from a tweet posted by the Alibaba Qianwen team on the X platform, announcing the open - sourcing of the Qwen 3.5 small - model series. There are four parameter specifications: 0.8B, 2B, 4B, and 9B.

Not long after the tweet was posted, a familiar name appeared in the comment section: Elon Musk.

He left a message: "Impressive intelligence density."

This is not the first time.

Around the Spring Festival, ByteDance released Seedance 2.0, and he retweeted and commented on it. Going further back, when Anthropic accused DeepSeek and Kimi, he was the first to jump out and retort, posting several sarcastic tweets and scolding more fiercely than anyone else.

Why does an American tech giant always support Chinese AI?

This seems abnormal, but if you lengthen the timeline and look at these events together, you'll find that Musk is waging a carefully - calculated multi - front war for his business empire.

1. Why are small models so exciting to Musk?

Let's first explain the term Musk praised.

Intelligence density, literally speaking, refers to how much intelligence can be packed into a limited number of parameters.

Among the four small models open - sourced by Qwen 3.5 this time, the smallest one has only 0.8B parameters. What does 800 million parameters mean? It's less than one - hundredth of GPT - 4. But it can have smooth conversations, understand complex instructions, and run on mobile phones and embedded devices.

Why is this so exciting to Musk?

Because Tesla's Optimus humanoid robot and FSD autonomous driving are extremely dependent on small models.

Imagine, where is the "brain" of the Optimus robot located? It's in its own body. It can't carry a data center around the streets, nor can it request data from the cloud every time it makes a move. It must complete visual recognition, path planning, and motion control locally, within milliseconds.

The same goes for FSD. Tesla has sold over 5 million cars, and each car is processing a massive amount of visual data in real - time. If every frame of the image needs to be uploaded to the cloud, the network bandwidth will be overwhelmed, and the latency will be a disaster.

So what does Tesla need? It needs small models that can run locally, have extremely low power consumption, extremely fast inference, but still maintain a certain level of intelligence.

The 0.8B and 2B models of Qwen 3.5 exactly meet this requirement.

What Musk praised is not Alibaba, but efficiency and common sense.

The current trend in Silicon Valley is brute - force computing power. People pile up tens of thousands of H100s to run a model with trillions of parameters. Whoever has more money is the boss. But Musk's philosophy has always been the first - principles thinking: solve the most fundamental problems with the least resources.

Musk once said something in a podcast that's worth pondering: Those who live in the software world haven't realized that they're about to take a big fall in the hardware field.

What does it mean?

Software can be infinitely large, but hardware has physical boundaries. The robot's head is only so big, the computing - power chip in a car has limited power consumption, and the battery capacity is limited. You can run a model with trillions of parameters in a data center, but you can't do it in the physical world.

And the physical world is Musk's ultimate battlefield.

2. Praising China, but worried about himself

After praising the models, let's talk about money.

Tesla's sales in the Chinese market account for more than one - third of its global sales, and its Shanghai factory is its largest production base. In 2026, Tesla plans to invest over $20 billion, focusing on areas such as AI computing power, robot factories, and the mass - production of the autonomous electric vehicle Cybercab.

China is an indispensable part of such a large - scale plan.

More direct evidence is that Tesla's Chinese branch is already using Chinese AI models. According to Bloomberg, Tesla plans to integrate DeepSeek and ByteDance's AI models into its in - car voice assistants.

Why not use its own Grok? The reason is embarrassing. After Grok was launched in the US market, it frequently made headlines due to issues such as anti - Semitic responses and sexualized images. Controversies overshadowed its technological progress.

Chinese consumers are not interested in Grok, at least compared to DeepSeek and other US AI models.

For Musk, this is not just a matter of face, but also a business issue. If Tesla can't provide the best AI experience in China, its sales will continue to decline.

So his praise of Chinese AI is an attempt to curry favor with the Chinese market.

But this is only the surface.

On a deeper level, it's because of the computing - power hunger of his own AI company, xAI.

xAI has just merged with SpaceX to form an entity valued at $1.25 trillion. The purpose of the merger is clear: SpaceX can provide xAI with computing power, talent, and data. SpaceX's satellite - internet service, Starlink, recently updated its privacy policy, clearly stating that it can collect user information for model training.

But even so, the computing - power scale of xAI is still far smaller than that of OpenAI, which has promised to invest over $1.4 trillion in data centers and chips.

Musk's praise of the computing - power potential of Chinese AI is actually putting pressure on the US domestic power - grid policy. He has criticized the aging US power grid and the slow approval process on multiple occasions, warning that the US may fall behind in the AI race.

Praising China is a way to criticize the US.

3. Criticizing opponents, using China as a weapon

On February 23rd, Anthropic published a strongly - worded blog post, naming DeepSeek, Moonshot, and MiniMax in the title and accusing them of "industrial - scale distillation" of Claude's core capabilities.

Who was the first to jump out and scold back? Elon Musk.

He posted several tweets on X, speaking very directly: How dare they steal what Anthropic stole from human programmers? It's an established fact that Anthropic stole a large amount of training data and paid billions of dollars in compensation.

This statement hits an embarrassing fact: Anthropic itself paid $1.5 billion in compensation for using pirated book training data last year.

But why is Musk so enthusiastic? Is it really just out of a sense of justice?

The answer lies in the timeline.

On February 16th, Axios reported that the Pentagon had warned Anthropic that it would pay a price and threatened to list it as a supply - chain risk. The root cause of the conflict is that Anthropic refused to allow Claude to be used for large - scale surveillance and the development of autonomous weapons.

On February 20th, news of the tense relationship between the two parties further fermented.

On February 23rd, it was reported that the Secretary of Defense would summon Anthropic's CEO, Dario Amodei, the next day. On the same day, it was revealed that Musk's xAI had signed an agreement with the Pentagon, allowing it to deploy Grok in specific systems.

It was also on the same day that Anthropic published that accusatory blog post.

What follows is interesting.

xAI had just snatched Anthropic's military contract, and the next day, Anthropic accused Chinese AI companies of "stealing technology," and these Chinese AI companies happen to be the ones Musk has been praising.

By scolding Anthropic, Musk can both strike at his competitors and build a good reputation for himself. One side is a hypocrite crying "thief," and the other side is a true hero speaking up for justice.

He also easily gained a wave of public favor.

4. Betting on the narrative, seeking the rules

In a program in January, Musk said something that attracted a lot of attention. He said that China would far surpass the rest of the world in the field of artificial - intelligence computing power.

As soon as he said this, there was a heated discussion in the Western AI circle. Some said that Musk was currying favor with China, while others said that this was an objective analysis based on the reality.

Musk's own reason is: The end - point of computing power is electricity.

Over 50% of the US power - grid equipment has been in operation for over 20 years. The interconnection capabilities of the three major power grids are weak, and electricity prices have started to rise. In contrast, China's power generation is twice that of the US, with a total installed capacity of 3.8 billion kilowatts and a very high redundancy.

This statement is becoming a consensus in the AI circle. Goldman Sachs also pointed out in a previous report that power shortages may slow down the US in the AI race.

Musk's praise of China is his way of sending a message: If the US doesn't reform soon, it will be too late.

But this is not all.

On February 27th, Trump signed an executive order, officially listing Anthropic as a national - security supply - chain risk and requiring all federal agencies to stop using its technology within six months.

The reason given by the Pentagon is that Anthropic refused to allow AI to be involved in fatal military decisions such as nuclear war. A serious conflict broke out during the discussion of "whether the military can use AI to assist in intercepting if the US is attacked by intercontinental ballistic missiles."

Anthropic's CEO, Amodei, responded strongly: "These threats from the Department of Defense won't change our stance. We can't in good conscience agree to their requests."

On the same day that Anthropic was "blacklisted," OpenAI and xAI quickly took action to fill the market vacuum.

Behind Musk's praise of Chinese AI and criticism of Anthropic, there is a deeper calculation: he wants to take control of the narrative of technological equal rights.

In his eyes, OpenAI and Anthropic are new monopolists trying to put AI in a cage and decide who can use it and who can't. Chinese AI, on the other hand, is the flag - bearer of technological equal rights, being open - source, affordable, and accessible to everyone.

This perfectly matches his long - standing persona: Anti - establishment, anti - monopoly, and anti - authority.

In this game, Chinese AI has indeed met his expectations.

[Beyond the Page] Words:

This is not the first time Musk has praised Chinese AI, and it won't be the last.

These three aspects combined form the complete picture of Musk.

The first aspect is business. Tesla wants to sell cars in China, and xAI needs computing power. Praising Chinese AI paves the way for himself. The second aspect is personal grudges. He has old scores with OpenAI and new hatreds with Anthropic. He uses Chinese AI as a weapon to deal with his enemies. The third aspect is the narrative. He wants to seize the flag of technological equal rights and place himself on the moral high - ground.

As for whether Chinese AI is really worthy of praise? That's another question.

This article is from the WeChat official account "Beyond the Page", author: Huahua. Republished by 36Kr with permission.